Orphaned Azure Database for PostgreSQL Flexible Server is billed daily but does not exist in ARM, Portal, CLI or Resource Graph (leftover of automatic Single Server migration)

Andreas Roth 20 Zuverlässigkeitspunkte
2026-09-02T09:08:59.85+00:00

We are billed every day for an Azure Database for PostgreSQL Flexible Server that does not exist anywhere in Azure Resource Manager. We cannot see it, stop it or delete it. We need a way to delete it, or someone at Microsoft to delete it, so the charges stop.

The resource, as reported by billing

  • Subscription: Pay-As-You-Go, ID ending in ...bbf3
  • Resource ID from usage details: /subscriptions/<sub>/resourceGroups/timetracking_esveo/providers/Microsoft.DBforPostgreSQL/flexibleservers/timetrackingdbesveo
  • Region: North Europe
  • Meters billed daily: B1MS Burstable compute (24 h/day), Flex Server Storage Data Stored, and since 1 September 2026 also "Extended Support vCore" (24 h/day), which alone is now about 2.30 EUR/day

Timeline reconstructed from Cost Management

  • Until 24 September 2024 the resource group timetracking_esveo contained a PostgreSQL Single Server (billed as vCore, Basic Data Stored, Backup LRS) and an App Service (S1).
  • On 17 September 2024 Flexible Server meters (B1MS, Storage Data Stored) started for the same resource group. This matches the automatic Single Server to Flexible Server migration Microsoft ran ahead of the Single Server retirement. Single Server meters stopped on 24 September 2024.
  • On 6 March 2025 the App Service meters stopped. This is when the resource group was deleted. The Flexible Server meters did not stop.
  • Since then the Flexible Server has been billed every single day (about 15 EUR/month, roughly 270 EUR so far) even though the resource group and the server are gone from ARM. Since 1 September 2026 the Extended Support meter has been added on top.

What we have verified

  • az group show -n timetracking_esveo returns ResourceGroupNotFound.
  • GET on the full resource ID above (api-version 2024-08-01) returns ResourceGroupNotFound. The same for the Single Server resource type.
  • az postgres flexible-server list on the subscription returns only our other server (esveo-shared-pg, West Europe).
  • az resource list --name timetrackingdbesveo returns nothing.
  • Azure Resource Graph across all subscriptions in the tenant (resources | where name =~ 'timetrackingdbesveo' or type has 'dbforpostgresql') returns only esveo-shared-pg.
  • The Microsoft.DBforPostgreSQL resource provider is registered on the subscription.
  • The server is not present in any other subscription in the tenant.
  • Signed in as a user with Owner on the subscription and Global Administrator in the tenant, and as the subscription's Account Administrator: same result. Portal "All resources" with hidden types shown does not list it either.
  • Cost analysis grouped by resource shows two PostgreSQL resources (esveo-shared-pg and timetrackingdbesveo); the Resources view shows only esveo-shared-pg.

Support history

We opened billing support request 2608170050001071. The Azure Billing team confirmed they cannot act on resource-level issues and suggested either cancelling the whole subscription (not an option, other production resources run in it) or buying a support plan to open a technical request. We do not think we should have to buy a support plan to stop paying for a resource that Microsoft's own automatic migration left behind and that has no ARM representation we could act on.

Questions

  1. Is there any customer-side way to delete a Flexible Server that exists in billing but whose resource group and ARM record are gone? Would recreating an empty resource group with the same name in the same region make it visible again?
  2. Is this a known issue with the automatic Single Server to Flexible Server migration, where deleting the resource group afterwards removes the ARM record but leaves the backend server running and billing?
  3. Can a Microsoft moderator route this to the Azure Database for PostgreSQL team or open a technical support case on our behalf, given that billing support has confirmed it is out of their scope?

Thank you.

Azure-Datenbank für PostgreSQL

1 Antwort

Sortieren nach: Älteste
  1. Smaran Thoomu 35,970 Zuverlässigkeitspunkte Externe Microsoft-Mitarbeiter Moderator
    2026-10-25T17:51:09.01+00:00

    Hey Andreas — that situation sounds really frustrating: you’re seeing daily charges for a PostgreSQL Flexible Server that isn’t discoverable anywhere in ARM/Portal/CLI/RG discovery (even though the backend is still billing). Based on the related Microsoft Q&A and the RM synchronization/refresh guidance, this most closely matches an Azure Resource Manager (control-plane) synchronization/metadata visibility issue, where the backend still exists but ARM/metadata is out of sync.

    What you can try (customer-side)

    1. Check for RM/backend sync/metadata refresh via Azure Resource Manager
      • Microsoft has described similar cases where a PostgreSQL resource was missing from portal/ARM/Resource Graph even though it was still active/billing.
      • In those cases, the resolution was an “ARM Backend Synchronization & Discovery Refresh” (i.e., an ARM metadata refresh from the backend) so the control plane can “re-bind” to the running resource.
    2. If the server is temporarily visible elsewhere, verify status and try deletion again
      • There’s also precedent that sometimes the resource “reappears” after an ARM cache/metadata refresh, and then deletion becomes possible normally.
      • The ARM refresh guidance also mentions checking for locks that could block deletion:
        • In the portal (if/when the server appears), look for Overview → Locks and ensure no “deleted”/blocking lock is preventing cleanup.
    3. If deletion was attempted and shows “ResourceNotFound”
      • Another Q&A thread indicates cases where the server becomes stuck in an intermediate lifecycle state (for example, dropping), after which deletion may need to be addressed once backend state is aligned.
      • The key takeaway: even if Portal/CLI say ResourceNotFound, the backend may still be in a lifecycle state and requires backend/ARM action to reconcile.

    What to request from Microsoft (the most likely path)

    Since Billing support confirmed they can’t act on resource-level issues, the most relevant escalation type is Azure Resource Manager metadata refresh (Technical request, service: Azure Resource Manager).

    The guidance in the RM sync article is essentially:

    • Create a new Help + Support request in the Azure portal
    • Issue type: Technical
    • Service: Azure Resource Manager
    • Category/selection: Resource management → other resource management issues
    • Ask specifically for an ARM metadata refresh / ARM backend synchronization & discovery refresh for the affected Flexible Server resource ID.

    Once ARM refresh completes, check whether the server becomes visible in:

    • Azure Portal (and then attempt deletion), and/or
    • Azure Resource Graph / Azure CLI

    This is consistent with the Q&A case where the server became visible again after the backend synchronization was completed.

    Notes about your specific scenario (Single Server migration leftover)

    Your hypothesis (automatic Single Server → Flexible Server migration leaving behind an orphaned billed instance) is plausible given that Single Server was retired and some migration windows involved enabling migration for a limited period. However, the provided documentation we have here doesn’t confirm a specific “orphaned flexible server after ARM cleanup” known issue pattern—it only provides the general retirement/migration context and the RM refresh remedies for visibility/orphan-like symptoms.

    Follow-up questions (to help an engineer craft the escalation and validate next steps)

    1. In Cost Management billing, does the Flexible Server billing stop anywhere after requesting an ARM refresh, or does it continue unchanged?
    2. What is the exact billing resource ID format you see in usage details (you already have one—confirm it’s the same ID you included in the support request)?
    3. Do you see any relevant entries in Activity Logs for the Flexible Server resource operation types around the time the resource group was deleted (e.g., deletes, provider operations, backend sync-related operations)?
    4. When you submit the support request, can you reference it as a control-plane sync problem and include that:
      • resource group returns ResourceGroupNotFound
      • ARM/Resource Graph/CLI do not list the server
      • billing continues
    5. (Optional but useful) Are you able to reproduce “missing from portal” by using a different role/account (just to rule out RBAC caching/permissions vs true absence)?

    Reference list (from the provided documentation)

    Hope this helps—if you share the support request number you already opened (in a way that doesn’t include sensitive info) and confirm whether you already tried an “Azure Resource Manager” technical request (as opposed to billing support), we can suggest the exact wording to use in the ticket.

    War diese Antwort hilfreich?

    0 Kommentare Keine Kommentare

Ihre Antwort

Antworten können von Fragestellenden als „Angenommen“ und von Moderierenden als „Empfohlen“ gekennzeichnet werden, wodurch Benutzende wissen, dass diese Antwort das Problem des Fragestellenden gelöst hat.